summer sea surface temperature conditions in ... - cerfacs.fr · 1september 2004 cassou et al. 3349...

1 SEPTEMBER 2004 3349C A S S O U E T A L .

q 2004 American Meteorological Society

Summer Sea Surface Temperature Conditions in the North Atlantic and Their Impactupon the Atmospheric Circulation in Early Winter

CHRISTOPHE CASSOU

Climate Global Dynamics Division, NCAR, Boulder, Colorado, and Climate Modelling and Global Change Team, SUC, CERFACS,Toulouse, France

CLARA DESER

Climate Global Dynamics Division, NCAR, Boulder, Colorado

LAURENT TERRAY

Climate Modelling and Global Change Team, SUC, CERFACS, Toulouse, France

JAMES W. HURRELL

Climate Global Dynamics Division, NCAR, Boulder, Colorado

MARIE DREVILLON

Climate Modelling and Global Change Team, SUC, CERFACS, Toulouse, France

(Manuscript received 26 February 2003, in final form 3 February 2004)

ABSTRACT

The origin of the so-called summer North Atlantic ‘‘Horseshoe’’ (HS) sea surface temperature (SST) modeof variability, which is statistically linked to the next winter’s North Atlantic Oscillation (NAO), is investigatedfrom data and experiments with the CCM3 atmospheric general circulation model (AGCM). Lagged observationalanalyses reveal a linkage between HS and anomalous rainfall in the vicinity of the Atlantic intertropical con-vergence zone. Prescribing the observed anomalous convection in the model generates forced atmospheric Rossbywaves that propagate into the North Atlantic sector. The accompanying perturbations in the surface turbulentand radiative fluxes are consistent with forcing the SST anomalies associated with HS. It is suggested that HScan therefore be interpreted as the remote footprint of tropical atmospheric changes.

The ARPEGE AGCM is then used to test if the persistence of HS SST anomalies from summer to late fallcan feed back to the atmosphere and have an impact on the next winter’s North Atlantic variability. ObservedHS SST patterns are imposed in the model from August to November. They generate a weak but coherent earlywinter response projecting onto the NAO and therefore reproduce the observed HS–NAO relationship obtainedfrom lagged statistics. Changes in the simulated upper-level jet are associated with the anomalous HS meridionalSST gradient and interact with synoptic eddy activity from October onward. The strength and position of thetransients as a function of seasons are hypothesized to be of central importance to explain the nature, timing,and sign of the model response.

In summary, the present study emphasizes the importance of summer oceanic and atmospheric conditions inboth the Tropics and extratropics, and their persistence into early winter for explaining part of the NAO’s low-frequency variability.

1. Introduction

Over the last couple of years, interest in the low-frequency variability of North Atlantic climate has in-creased markedly (e.g., Marshall et al. 2001; Stephenson

Corresponding author address: Dr. Christophe Cassou, CERFACS-SUC (URA 1875), Climate Modelling and Global Change Team, 42Ave. Gustave Coriolis, 31057 Toulouse, France.E-mail: [email protected]

et al. 2003). In particular, attention has been focused onthe North Atlantic Oscillation (NAO), described as themost prominent and recurrent pattern of atmosphericvariability (Wallace and Gutzler 1981). Important eco-logical, economical, and societal impacts that the NAOexerts on the entire Northern Hemisphere have beendocumented and questions have been raised about itspotential predictability (Hurrell 2003). Although theNAO mainly arises from internal atmospheric interac-

3350 VOLUME 17J O U R N A L O F C L I M A T E

tions, external forcings such as sea surface temperature(SST) anomalies or increasing greenhouse gases, etc.might affect the occurrence and/or the persistence of itsphase. Their influence could be important for under-standing both amplitude and long-term evolution of theNAO in the context of climate change or climate pre-dictability.

On one hand, the impact of atmospheric circulationanomalies upon extratropical/subtropical oceans hasbeen demonstrated in numerous studies (see, e.g., Bat-tisti et al. 1995). Anomalous turbulent heat fluxes, buoy-ancy-driven entrainment and Ekman advection are themain processes whereby the atmosphere imprints large-scale SST anomalies. In such a scenario, the extratrop-ical atmosphere primarily leads the midlatitude oceanvariability by a month or so, as examined for instancein Deser and Timlin (1997) or Frankignoul et al. (1998).The low frequency oceanic anomalies are considered asthe passive temporal integration of the atmospheric sto-chastic forcing (Frankignoul and Hasselmann 1977).

On the other hand, although the atmospheric fluctu-ations are undoubtly the dominant driver of the upper-ocean anomalies, the role of the ocean has been sug-gested to explain the reddening of the atmospheric fluc-tuations from seasonal to decadal time scales. For in-stance, Rodwell et al. (1999) and Mehta et al. (2000)using atmospheric global circulation models (AGCMs)forced by historical global SST estimations have beenable to reproduce some of the low-frequency variabilityof the winter NAO during the last 50 years. The dom-inant influence of the tropical/subtropical Atlantic oce-anic basin has been inferred in some studies (e.g., Suttonet al. 2001). Terray and Cassou (2002) diagnose frommodel simulations that anomalous SSTs there alter thelocal atmospheric Hadley cell over the western tropicalbasin, providing an anomalous source of Rossby waves(Sardeshmukh and Hoskins 1988) extending northeast-ward from the Caribbean toward Europe. The phase ofthe NAO is ultimately affected by this teleconnectionmechanism as described in Peng et al. (2003), Drevillonet al. (2003), and Cassou et al. (2004). Hoerling et al.(2001) argued that the whole Tropics, not just the At-lantic domain, must also be considered. In particular,the role of the Indian Ocean has been suggested bySutton and Hodson (2003) and Hurrell et al. (2004).

North Atlantic midlatitude SST anomalies may alsohave a nonnegligible influence on the variability of theNAO. The dominant mechanisms are discussed in Pengand Whitaker (1999) and Peng and Robinson (2001).They emphasize the role of synoptic storm activity andassociated eddy–mean flow interaction that can be per-turbed by the existence of anomalous meridional SSTgradients along the subpolar–subtropical gyre front. Ex-tratropical oceanic anomalies may also contribute to thereddening of the NAO spectrum in the interannual bandthrough the reemergence mechanism in which upper-ocean temperature anomalies created over a deep mixedlayer in winter are preserved within the highly stratified

summer thermocline (Deser et al. 2003). As they re-appear at the surface at the next late fall or early winterby reentrainment (de Coetlogon and Frankignoul 2003),they may have a feedback impact upon the overlyingatmosphere. However, conclusions about extratropicalSST forcing upon the North Atlantic atmosphere arevery model dependent [see Robinson (2000) or Kushniret al. (2002) for a review] and should be taken withcaution. The models’ different sensitivities appear to bedue to a different representation of the simulated cli-matological flow and internal atmospheric modes of var-iability.

At monthly to seasonal time scales, lagged relation-ships between the early wintertime North Atlantic at-mosphere and the previous summer–fall oceanic con-ditions have been documented by Czaja and Frankignoul(1999, 2002) and Drevillon et al. (2001). Using singularvalue decomposition (SVD), a statistical link is extract-ed between the so-called summer North Atlantic‘‘Horseshoe’’ (HS) SST pattern and the NAO, as illus-trated in Fig. 1 using data from the National Centersfor Environmental Prediction–National Center for At-mospheric Research (NCEP–NCAR) reanalysis (Kalnayet al. 1996). The negative phase of the NAO (Fig. 1a)in early winter [October–December (OND) mean] ischaracterized by a slackening of the Icelandic low andthe Azores high, and is associated with late summer[August–September (AS) mean] negative SST anoma-lies located southeast of Newfoundland and extendingwestward slightly along the Gulf Stream path, surround-ed by positive anomalies on the eastern side of the At-lantic (Fig. 1b). Maximum warming occurs off thenorthern African coast and from Britain to the southerntip of Greenland. The HS mode is dominated by inter-annual fluctuations with a tendency toward dominantpositive phases in the 1950s and 1960s and more re-current negative phase from the early 1970s to the early1990s (Fig. 1c). Watanabe and Kimoto (2000) and Dre-villon et al. (2001, 2003), from data and model exper-iments respectively, speculate on physical processes re-sponsible for the statistical lagged properties shown inFig. 1, but the nature of the summer ocean influence onthe next winter NAO is still unclear.

The main goal of this paper is to go beyond statisticalanalyses and to suggest a physical explanation for theexistence of the summer HS pattern and a hypothesisfor the lagged relationship between that SST pattern andthe following winter atmospheric conditions. The ob-servational results are used to motivate several modelexperiments to test these ideas. The paper is organizedas follows. A combined observation–model analysis ispresented in section 2 to illustrate the relationship be-tween anomalous summer atmospheric conditions overthe Atlantic and the oceanic HS mode. Interpretationsabout the origin of the HS are discussed. The laggedinfluence exerted by the HS on the North Atlantic win-tertime atmosphere is then examined in section 3. Theobserved HS monthly evolution is prescribed from late


FIG. 1. First singular value decomposition modes calculated be-tween (a) OND SLP (hPa) and (b) AS SST (8C) computed fromNCEP–NCAR over the 1949–2001 period. The covariance betweenthe two modes is equal to 63.9%. Contour intervals are respectively0.1 hPa and 0.18C. (c) Normalized time series related to the AS SST(bars) and OND SLP (solid line) leading SVD mode. The correlationbetween the two PC time series reaches 0.72 (significant at 99%).

summer to early winter in a model to investigate itsimpacts on the North Atlantic atmospheric dynamics. Asynthesis and discussion are provided in section 4.

2. Origin of the late summer North Atlantic‘‘Horseshoe’’ SST mode

a. Motivation: Observational results

The origin of the late summer HS mode that exhibitsthe strongest covariability with the next winter NAO

(Fig. 1) is investigated first from observations. To clear-ly identify the role of the atmosphere in the origin ofthe ocean anomalous pattern, June–July (JJ) averagedatmospheric fields have been regressed onto the AS av-eraged SST SVD principal component (PC) time series.All atmospheric fields have been quadratically detrend-ed prior to the regression.

There is some evidence that the extratropical HSmode is forced in part by atmospheric circulation anom-alies through modifications of the surface turbulent andradiative fluxes. The lagged regression of the total sur-face heat flux estimated from the ComprehensiveOcean–Atmosphere Data Set (COADS; Woodruff et al.1998) following the Cayan (1992) formulation matchesthe spatial pattern of the HS SST anomalies fairly well,with the atmosphere tending to warm the subtropicaland eastern Atlantic and to cool the western basin (Fig.2a).

The turbulent flux contribution can be mainly un-derstood in terms of near-surface wind changes (Fig.2b). The anomalous cyclonic circulation located offNewfoundland induces anomalous northerlies over thewestern part of the basin, slackened westerlies at higherlatitudes especially in the Labrador and Irminger Seas,and reduced trade winds within the subtropics in thecentral Atlantic. Cold air is advected from the north onthe western side of the basin, cooling the ocean mixedlayer, while decreased winds within the band 508–658Nreduce the turbulent heat flux, warming the mixed layer.Besides, Ekman pumping due to the anomalous cycloniccirculation over the basin may contribute to coolingdown the ocean surface off Newfoundland (not shown).Advection is also expected to explain part of the anom-alies along the Gulf Stream path where local wind anom-alies are not clearly related to surface oceanic condi-tions.

The trade winds are not significantly modified in theeastern subtropical basin where the maximum warmingoccurs. Over this specific region, the radiative balanceat the surface might have a complementary or evengreater role. Decreased cloudiness is observed on theeastern side of the Atlantic basin especially within thesubtropical band and off Europe (Fig. 2c), consistentwith the higher SSTs. Reduced low-level clouds that aredominant over this area diminish the so-called albedofeedback and tend to warm up the surface ocean. In-creased cloudiness over most of the anomalous SST coldcore reinforces the cooling of the ocean mixed layer dueto enhanced turbulent heat loss. Although less signifi-cant and less large scale, the NCEP–NCAR surfaceshortwave flux anomalies match the COADS cloudinessresults. The cloud pattern is persistent until fall (notshown) and may contribute to maintain the ocean sur-face anomalies. This is particularly true above the sub-tropical SST lobe. The western tropical Atlantic basin(west of 608W) is characterized by increased cloudinessover positive SST anomalies. Low-level stratiformclouds are fairly sparse over this area dominated by


FIG. 2. Two-month lead (JJ) regression onto the AS SST PC of (a)total surface heat flux (W m22) from COADS and (b) 1000-hPaNCEP–NCAR wind superimposed on the SVD SST mode itself. Con-tour intervals are 0.18C. (c) Cloudiness (octas) from COADS and (d)OLR (W m22) (contour) from NCEP–NCAR and rainfall station-based data (mm day21) from Hulme and New (1997) (contour intervalis 1 W m22 for OLR and positive values are indicative of decreaseconvection). Only points satisfying the 90% level confidence are plot-ted for all the datasets.

convection, and this pattern suggests some strengthen-ing of the convective activity.

We investigate this signal further using NCEP–NCARoutgoing longwave radiation (OLR) as a proxy for deepconvection in the Tropics. As accurate measurementshave been only available from satellites since the late1970s, station-based precipitation data (Hulme and New1997) are used to further confirm the robustness of ourfinding from OLR. The HS mode is associated with earlysummer enhanced convective activity (decreased OLR)over Africa and over the Caribbean Sea extending north-westward toward Florida (Fig. 2d). Such an increase ofconvection within the band 108–208N, indicative of aslight northward shift and intensification of the inter-tropical convergence zone (ITCZ), contrasts with mostlydry conditions over the Amazonian basin (increasedOLR). The precipitation station data confirm the OLRsignal over land (Fig. 2d) and are consistent with in-creased cloudiness over the Caribbean diagnosed in-dependently from COADS. The anomalous rainfall pat-tern shown in Fig. 2d is similar to that documented byGiannini et al. (2003) at interdecadal time scale and bySutton et al. (2000) and Okumura et al. (2001) at in-terannual time scales. As shown by Fontaine et al.(1999), when the SST anomalies are positive in thenorthern tropical Atlantic, the local Hadley circulationis altered with increased low-level convergence fromGulf of Guinea to the Sahel bringing more moistureover land consistent with enhanced convection. By con-trast, mostly dry conditions dominate along the equatorover sea and over the South American continent.

b. CCM3 model experiments

We have shown from observations that the late sum-mer North Atlantic SST anomalies associated with theHS mode are likely a lagged local response to the ex-tratropical underlying atmospheric changes. We havealso shown that HS is preceeded by anomalous tropicalatmospheric anomalies. To explore the extent to whichtropical Atlantic rainfall changes can affect the midlat-itude atmosphere in summer and how these are linkedto HS, two experiments are conducted with the thirdversion of the NCAR Community Climate Model(CCM3), which has a T42 triangular horizontal reso-lution and 18 vertical levels. A complete description ofthe physical and numerical methods used in CCM3 isprovided in Kiehl et al. (1996) and model validation isgiven in Hurrell et al. (1998). The summer ITCZ in-tensity is artificially perturbed in the model based uponthe observed anomalous convection pattern associatedwith HS (Fig. 2d). Two idealized heat sources have beenspecified, one centered over the Caribbean Sea and oneover the Sahel region (rectangles in Fig. 3a). The ver-tical distribution of the heating perturbation (Fig. 3b)reaches a maximum intensity at 500 hPa and is equalto 128C day21 , which roughly corresponds to twicethe maximum observed changes estimated from the


FIG. 3. (a) The JJ 2-month-averaged difference for OLR(W m22) between anomalous prescribed diabatic ‘‘heating’’ and dia-batic ‘‘cooling’’ CCM3 experiments. Contour interval is 10 W m22

and shading stands for positive anomalies. Convention is negativeOLR for increased convection. The thick solid boxes represent thegeographical domains where the anomalous diabatic heating/coolingis prescribed in CCM3. (b) Anomalous diabatic heating vertical pro-file (8C day21) for a section at 158N as a function of longitude andprescribed in CCM3. Contour interval is 0.58C day21.

FIG. 4. As in Fig. 3a, but for CCM3 (a) wind at 1000 hPa, (b)geopotential height at 500 hPa (m) and (c) total heat flux (W m22).Contour intervals are 5 m and 10 W m22. Shaded areas exceed the95% significance limit using T statistics.

NCEP–NCAR reanalyses. A twin experiment with aprescribed 228C day21 is also carried out and the modelis integrated for 10 summers in both cases. The readeris invited to refer to Branstator and Haupt (1998) for acomplete description of the experimental setup. Thechoice for the strong magnitude of the perturbation ap-plied in the model is dictated by our attempt to betterextract the tropical impact onto the midlatitude dynam-ics within the constraint of limited computing resources.In the following, we focus our explanation on the linear[(heating 2 cooling)/2] response.

As expected, the model shows enhanced convectiveactivity collocated with the prescribed anomalous dia-batic heating (Fig. 3a). Maximum OLR anomalies reachabout 260 W m22 centered over the Sahel and Carib-bean. Diminished convection appears over the SouthAmerican continent and the western equatorial Atlantic,reminiscent of the observations. Thus, the diminishedconvection may be interpreted as a remote response tothe increased convection along the ITCZ in the model.We have also verified that specifying a negative heatsource over Amazonia in the model leads to enhancedconvection over the Caribbean and also to a lesser extent

over the Sahel (not shown). This suggests that the fullheating distribution in the whole tropical Atlantic may,in fact, play a role. Very similar conclusions are ob-tained from the upper-tropospheric convergence field(not shown).

The simulated low-level wind response to the heatsources is shown in Fig. 4a. Weakened trade winds dom-inate the subtropical North Atlantic basin, consistentwith observations. Similarities can be also found be-tween the modeled and observed surface wind changesat midlatitudes. Despite a slight westward and northward


shift, the model is able to reproduce the cyclonic at-mospheric circulation off Newfoundland in associationwith a stronger Atlantic ITCZ. The perturbed atmo-spheric circulation is reminiscent of a Rossby wave trainpattern originating from the Gulf of Mexico and ex-tending northeastward toward Scandinavia. Althoughsuch a mechanism is more active in winter–spring (Web-ster 1982), it seems to be at work also in summer inthe model. It clearly dominates the remote midlatitudeatmospheric changes as displayed in Fig. 4b for the 500-hPa geopotential height (hereafter Z500). MinimumZ500 is located around Newfoundland and slightly westof the low-level cyclonic conditions while high pressureover Britain is consistent with the surface anticyclone.Note that the latter is less evident in the observations(Fig. 2b). The overestimation of the Rossby wave ex-tension simulated in CCM3 might be attributed to thehigher intensity of the perturbation prescribed in theTropics compared to observations or to the model over-sensitivity for that particular season due to mean statebiases. To measure the importance of the remote signal,we note that the signal-to-noise ratio comparing forcedvariability (model response in sensitivity experiments)and internal variability (estimated from a control sim-ulation of CCM3) is close to 1 off Newfoundland anddrops to 0.6 over the United Kingdom.

To further confirm the teleconnection hypothesis andits origin, complementary simulations have been carriedout by selecting only one of the two anomalous diabaticheating sources, thereby isolating the Sahel versus Ca-ribbean impacts. Similarly to the previous experiments,the model is integrated for 10 summers and for the‘‘heating’’ and the ‘‘cooling’’ cases. The Rossby-wavetype anomalous circulation completely disappears in theSahel-only experiment while a clear cyclonic circulationcentered off the North African coast (308N, 208W) dom-inates (not shown). In that case, the slackening of theAtlantic trade winds is more pronounced since the low-level convergence to the Caribbean has been removed.By contrast, Caribbean-only experiments (not shown)reveal a clear Rossby wave train while the trade windsare reinforced in the western tropical basin but not af-fected in the eastern basin. This result is consistent withthe fact that the western convection core is bordered onits northern side by climatological upper-level wester-lies, whereas the African core is capped by easterlies,precluding any meridional propagation of tropicalforced Rossby waves (Hoskins and Pierce 1983). Basedon these model results, we suggest that the observedAtlantic wind pattern associated with the HS in summercan be viewed as a remote response to changes in trop-ical convection over both the Sahel and Caribbean re-gions, with the trade winds responding to both heatsources and the extratropical winds responding mainlyto the Caribbean one.

Total surface heat flux anomalies over the subtropical/North Atlantic basin in the CCM3 experiments representthe local atmospheric forcing of the ocean mixed layer

(Fig. 4c). The heat flux anomalies are consistent withanomalous low-level wind conditions and bear a strongresemblance to the observed pattern (Fig. 2a). Weakertrade winds are associated with diminished turbulentheat flux out of the ocean east of 508W, which tends towarm up the subtropical North Atlantic. A secondarycore centered around 508N off Europe corresponds tothe midlatitude slackening of the mean westerlies andsimilarly tends to warm up the northeast Atlantic. Bycontrast, the western side of the basin north of 208N isdominated by enhanced evaporation and northerlywinds that tend to cool down the surface ocean, althoughsignals are rather weak and mostly confined south ofNewfoundland close to the Gulf Stream. Decomposingthe total heat flux into radiative and turbulent compo-nents (not shown) reveals that the latter dominates inthe model except off the Iberian penisula where thesimulated warming tendency appears to be limited tothe coast (Fig. 4c) compared to observations (Fig. 2a).This confinement can be attributed to an overestimationof enhanced cloudiness in the model over this area (theCCM3 cloud response to the perturbed ITCZ is maxi-mum to the east of the anomalous cyclonic circulationwhile in reality these are collocated, Fig. 2c). Exceptover this specific area, changes in radiation balance dueto cloud cover are weak in the model and are suspectedto be underestimated. This could be explained by theabsence of local ocean feedback to the atmosphere dueto the experimental configuration. In reality, local SSTanomalies have been shown within the subtropics tolocally affect the stability of the atmospheric columnand thus amplify and/or maintain the anomalous cloudpattern (Norris et al. 1998). The weakness of the cloudimpact could also be explained by the underestimationof low-level cloud amount in CCM3. Additional ex-periments using a newer version of CCM3 coupled toa complex ocean mixed layer model (Alexander et al.2002) are currently in progress to examine the SST re-sponse to tropical diabatic heating anomalies.

3. Summer SST forcing of the NAO in earlywinter

a. Experimental setup

The speculated feedback of SST anomalies associatedwith the HS mode onto the early winter atmosphere overthe North Atlantic–European sector is now examinedthrough two 30-member ensemble integrations, each 9months long (July–March), using the Action de Re-cherche Petite Echelle Grand Echelle (ARPEGE) modeljointly developed by Meteo-France and European Cen-tre of Medium-Range Weather Forecasts (Deque et al.1994). The present climate version of ARPEGE uses aT63 triangular horizontal truncation. Diabatic fluxes andnonlinear terms are calculated on a gaussian grid ofabout 2.88 latitude 3 2.88 longitude. The vertical isdiscretized over 31 levels (20 levels in the troposphere)


FIG. 5. Anomalous SST patterns (8C) added to (subtracted from) ARPEGE boundary conditions from (a) Aug to (d) Nov for the HS1(HS2) experiment. Contour interval is 0.28C and shading stands for positive anomalies.

using a progressive vertical hybrid coordinate extendingfrom the ground up to about 34 km (7.35 hPa). Furtherdetails about the model physics package and about thevalidation of the simulated intraseasonal and interannualvariability can be found in Doblas-Reyes et al. (1998,2001) and in Cassou and Terray (2001).

The HS-type SST anomalies are either added to (here-after HS1 experiment) or subtracted from (HS2) themodel boundary conditions from 15 July to 15 Decem-ber (Fig. 5), while late December to March oceanicconditions remain unperturbed and fixed to the seasonalclimatology. The monthly anomalous SST patterns havebeen obtained by regression of monthly SST data uponthe PC time series of the leading SVD mode shown inFig. 1c. The intensity of the SST anomalies imposed inthe model is given by the strongest value of the PC timeseries over the analyzed 52-yr period. For HS1, max-imum cooling off Newfoundland reaches 21.28C in Au-gust whereas subtropical values are slightly above10.68C off the northwest African coast. Maximumwarming peaks at about 118C off western Europe. Theopposite is taken for HS2.

As described in Drevillon et al. (2001), the HS oce-anic anomalies are fairly persistent from summer to ear-

ly winter (the correlation between August and Novem-ber PC time series calculated from empirical orthogonalfunction decomposition of the consecutive July–August–September–October–November normalized months andcapturing HS is equal to 0.44). Consistent with theirstudy, the extratropical lobes prescribed in our case ex-hibit a slow decay in amplitude with time (about 50%diminution from August to November), while a slighteastward propagation of the midlatitude SST core isnoticeable. Compared to the extratropics, the subtropicaloceanic anomalies are more persistent both in amplitudeand space. The rate of decay of the tropical/subtropicalSST anomalies are roughly consistent with the sto-chastic climate model (Frankignoul and Hasselmann1977). In the following, we examine the linear [(^HS1&2 ^HS2&)/2] component of model response (anglebrackets standing for ensemble mean).

b. ARPEGE atmospheric response to HS SSTanomalies

Zonal wind anomalies at 200 hPa (U200), Z500, andgeopotential height anomalies at 1000 hPa (Z1000) arepresented for the OND average for ARPEGE (Figs.


FIG. 6. Regression of NCEP–NCAR OND zonal wind at (a) 200 hPa (U200) (m s21), (b) geopotential height at 500 hPa (Z500) (m), and(c) at 1000 hPa (Z1000) (m), onto AS SST time series obtained from SVD. Linear ARPEGE response in OND for (d) U200, (e) Z500, and(f ) Z1000. Contour intervals are 0.5 m s21 for U200 and 2 m for geopotential. Shaded areas exceed the 95% significance limit usingT statistics.

6d,e,f) and observations (Figs. 6a,b,c). The latter areestimated from NCEP–NCAR and obtained by laggedregression onto the August–September SST SVD PCtime series. The model develops a global quasi-baro-tropic response to the prescribed HS SST anomaliesand reproduces fairly well the observed early winteratmospheric changes. The U200 zonally elongated tri-pole from the Tropics to high latitudes is well captured,especially on the eastern side of the basin. It corre-sponds to the reinforcement and eastward extension ofthe climatological subtropical upper-level jet. TheZ500 stationary waves are accordingly modified. Aclear barotropic negative NAO signal is extracted for

NCEP–NCAR (Figs. 6b,c), and the Z1000 anomaliesare very similar in location to the MSLP structure ex-tracted by SVD (Fig. 1a). ARPEGE geopotential heightresponses (Figs. 6e,f) bear some resemblance to theobserved patterns, with generally negative (positive)anomalies south (north) of 608N. However, the strengthand eastward extension of the positive anomalies near608–708N are significantly underestimated in the modeland the negative anomalies over the Barents Sea areopposite to observations.

A different picture emerges for the late summer (Au-gust–September) anomalies simulated by ARPEGE(Fig. 7). The Z500 NCEP–NCAR anomalies reveal a


FIG. 7. As in Figs. 6b,c,d,f, but for the Aug–Sep mean.

clear wavy pattern emanating from the western tropicalAtlantic basin and extending toward Scandinavia. Thispattern bears a strong resemblance to the one obtainedfrom the CCM3 sensitivity experiments as analyzed inthe previous section for early summer. This result sug-gests that the tropical forcing of the midlatitude atmo-sphere is persistent over the entire season and thereforetends to reinforce the HS-type oceanic pattern initiatedat the begining of summertime. ARPEGE results con-trast strongly. The model does not capture the forcedRossby wave pattern at 500 hPa but departures betweenobserved and modeled signals are even more strikingin the lower troposphere. While a clear quasi-barotropicstructure is extracted from NCEP–NCAR north of 358N,the model develops a baroclinic response with high pres-sure anomalies at the surface over the cold waters offNewfoundland and low pressure anomalies off Irelandover warm SST conditions. The model’s low-level re-sponse over the ocean is thus opposite to that fromNCEP–NCAR. South of 308N, both model and obser-vation exhibit a baroclinic profile in agreement with theclassical tropical/subtropical ocean–atmosphere rela-tionship (Hoskins 1987). At higher latitudes (north of658N), the barotropic properties of the atmospheric floware recovered in the model.

A couple of hypotheses may explain NCEP–NCAR/ARPEGE disagreements for late summer months at mid-latitudes. The first one would refer to the nature of theexperimental design, which could artificially perturb the

low-level atmospheric response. It is well known thatthe so-called Atmospheric Model Intercomparison Proj-ect (AMIP) type simulations do not correctly capturethe surface ocean–atmosphere interactions as developedin Barsugli and Battisti (1998) and may affect the baro-tropic shape of the atmospheric modes that dominatesat midlatitudes. Such a mechanism might be seasonallydependant and could dominate in ARPEGE in sum-mertime. A second hypothesis would be to consider themodel as realistic and to assume that it correctly re-sponds to the extratropical SST anomalies. The anom-alous low-level circulation in the model then would tendto damp the surface ocean anomalies and would be con-sistent with the idea that the extratropical atmosphereis not responsible for the genesis of the extratropicalanomalous HS lobes. Such a hypothesis reinforces thetropical origin of the anomalous extratropical atmo-spheric circulation suggested previously. It has been ver-ified that no significant changes are simulated in AR-PEGE within the tropical band. In particular, the AtlanticITCZ is neither affected in intensity nor in position (notshown). In such a scenario, the switch between summerbaroclinic and fall quasi-barotropic model responsewould also be considered as a realistic physical mech-anism. The HS anomalies would then move from a pas-sive role in summer to an active role in winter in theforcing of the extratropical atmosphere. In the rest ofthe paper, we further document such a hypothesis andfocus our interest on the origin of the simulated baro-


FIG. 8. Anomalous vertical profile for (a) zonal wind and (b) geopotential height for Sep–Nov months and averagedover the 308–408N, 508–108W domain. Units are respectively (m s21) and (m).

tropic response in fall. This will allow us to providesome possible explanations about the observed link be-tween the phase of the summer oceanic HS mode andthe early winter atmosphere.

c. Role of the climatological background in thesimulated atmospheric response

Figure 8 shows anomalous vertical profiles of zonalwind and geopotential height in September, October, andNovember for the region (308–408N, 508–108W), whichcorresponds to the diffluence zone of the upper-tropo-spheric westerly wind jet and encompasses the maxi-mum U200 westerly anomaly simulated by the model(Fig. 6d). While all three months exhibit positive ver-tical wind shear anomalies, the shear is amplified con-siderably in October and November compared to Sep-tember (Fig. 8a). The vertical structure of the geopo-tential height response in the months with strong verticalwind shear is approximately equivalent barotropicwhereas it is baroclinic in the months with weak shear(Fig. 8b). Why do similar SST anomaly patterns leadto different vertical structures of atmospheric circulationresponse in the presence of vertical wind shears?

Some studies suggest that anomalous transients andassociated feedbacks onto the mean flow play an im-portant role in shaping the midlatitude atmospheric re-sponse to extratropical SST anomalies (Held et al. 1989;Watanabe and Kimoto 1999). The relationship betweensynoptic-scale eddies along the midlatitude storm trackand local vertical shear has been extensively docu-mented in the literature (see, e.g., Hoskins and Valdez1990). The model anomalous eddy activity estimatedby the variance of the bandpass (2.2–6 days) filteredZ500 is presented in Fig. 9 as a function of latitude andfor the same longitudinal band used in Fig. 8, for themonths September, October, and November. The month-ly U200 simulated anomalies are also shown. In Sep-

tember, although U200 westerly anomalies dominatesignificantly between 308 and 508N, the storm activity(hereafter STA) is not significantly altered. In Octoberhowever, westerly U200 anomalies between 308 and458N are accompanying by a significant enhancementin STA. In November, despite a slight 58 misfit, en-hanced (reduced) STA coincides rather well in locationover the entire eastern basin with U200 anomalous ac-celeration (slackening) and is overall amplified. Thus,it appears that increased vertical wind shear is accom-panied by enhanced eddy activity changes from Octoberonward suggesting, all together, a modification of thebaroclinicity.

The alteration of the transients as well as their sea-sonality are presumably of central importance to explainthe nature of the mean atmospheric response (Peng etal. 1995). Climatological monthly mean tendencies inU200 and STA computed as the difference between thegiven month and the previous one for the same region(308–408N, 508–108W) are shown in Fig. 9d. In Sep-tember the mean STA is at its minimum amplitude,while the upper-level climatological jet already startsincreasing over the selected area. It is hypothesized herethat the baroclinic instability criteria are not met yet forSeptember at this location. The jet might not be strongenough to trigger changes in storminess, despite locallyenhanced vertical shear (Fig. 8a) interpreted as a directresponse to the anomalous meridional temperature gra-dient imposed at the surface. From October onwardU200 and STA are intensifying with maximum strength-ening in November for the latter, one month later forthe former. This crude analysis tends to indicate that thedevelopment of synoptic storms in the model requiresa minimum speed of the jet. When such a threshold isovertaken, storms rapidly reach their maturity and donot follow as closely the gradual increase of the upper-level westerly circulation like in December. Due to amore favorable climatological background starting in


FIG. 9. Anomalous zonal wind at 200 hPa (U200) and storm track activity (STA) as a function of latitude andaveraged over the longitudinal 408–108W band for (a) Sep, (b) Oct, and (c) Nov. Units are m s21 and m2. Thicksuperimposed lines exceed the 95% significance limit using T statistics. (d) Monthly changes for climatological U200and STA, estimated as the difference between the mean value of the considered field for a given month and the onefor a month earlier.

October, it is suspected, in our case, that the presenceof local vertical shear, in relation to local prescribedSST anomalies, is now able to affect the local baro-clinicity and related storminess at midlatitudes. It is thussuspected that the action of the storm explains to a greatextent the barotropic nature of the mean atmosphericresponse from October. This hypothesis relies on theproperties of the synoptic eddies to favor quasi-baro-tropic profiles as detailed, for instance, in Hoskins etal. (1983).

Beside its impact for explaining the nature of theatmospheric response, the alteration of the midlatitudeSTA may provide a strong positive feedback in main-taining and intensifying the modified mean flow. It isexpected to affect the planetary-scale waves and to mod-ify the stationary wave activity allowing for large-scale

anomalous circulation. The latter clearly appears in No-vember in the model as detailed below and is clearlyamplified in December (not shown). The main stormtrack is significantly reduced over the entire North At-lantic from the Labrador Sea to the British Isles, whileits southern edge is reinforced within a 308–458N lati-tudinal band (Fig. 10a). Such a pattern is consistent withthe negative phase of the NAO as described in Hurrell(2003). The simulated pattern bears some resemblancewith NCEP–NCAR changes estimated by Novemberlagged regression of the bandpass-filtered Z500 on theAS SVD PC time series (Fig. 10c), especially east of408W. ARPEGE tends, however, to underestimate thealteration of the synoptic eddies on the western side ofthe basin (Newfoundland area) at midlatitudes and tooverestimate the storminess reduction farther north.


FIG. 10. Linear ARPEGE model response in Nov for the (a) STA (m2) and (b) for the Eliassen–Palm (EP) flux. The mean model zonalwind at 200 hPa (U200) climatology is superimposed (m s21; thick solid line) and contours start at 20 m s21, every 10 m s21. Regressionof NCEP–NCAR Nov (c) STA and (d) EP onto AS SST time series obtained from SVD. Contour intervals for STA are 0.5 m 2 and shadedareas exceed the 95% significance limit using T statistics. The mean NCEP–NCAR U200 (thick solid line) is superimposed (d).

Such a discrepancy can be attributed to the mean biasof the model (position and strength of the storm trackand the jet) and the reader is invited to refer to Cassouand Terray (2001) or Drevillon et al. (2003) for furtherdetails.

Anomalous momentum fluxes associated with syn-optic eddies are presented in Figs. 10b,d for the modeland NCEP–NCAR observations using the Eliassen–Palm (EP) E vector following Trenberth’s (1986) defi-nition. The zonal E component can be interpreted as thezonal stretching of the mean flow induced by the syn-optic eddies and its meridional component as the dom-inant eddy forcing of the mean streamfunction evolution(Held et al. 1989). The divergence (convergence) of Eis indicative of mean flow acceleration (deceleration)due to the presence of STA changes. A clear divergenceoccurs in the model on the southeastern side of the meanclimatological jet (around 308N and within 408–108W)and is compensated by a strong convergence at its tailend off western Europe and at the entrance of the Af-rican subtropical jet. Such an anomaly is consistent withenhanced zonality of the basic flow and diminution ofthe planetary wave activity (Doblas-Reyes et al. 2001).It is also consistent with the forcing tendency due tosynoptic eddies to develop a negative NAO-type large-scale pattern, as the strength of the Azores high (Ice-landic low) is affected by the change in the main course

of the storms. Despite a clear 108 displacement to thenorth that can be related again to the model mean bias,a similar pattern can be found in NCEP–NCAR. TheEP divergence (convergence) is maximum along thesouthern (northern) edge of the mean jet over the easternbasin (east of 408W).

4. Conclusions

a. Synthesis

Seasonal ocean–atmosphere interaction in the Atlan-tic has been investigated to obtain a better understandingof the statistical relationship between the summer SSTanomalies associated with the ‘‘Horseshoe’’ mode andthe NAO in the following early winter. A combinedobservational and modeling approach was adopted toexamine the origin of the HS pattern and to providesome physical hypotheses for the delayed HS forcingon the next winter atmospheric circulation.

The results collectively suggest that HS can be pri-marily considered as an extratropical oceanic footprintof tropical convection changes in the Atlantic ITCZ viaatmospheric teleconnection mechanisms. When the cli-matological ITCZ is enhanced or displaced northward,the atmospheric circulation is characterized by a cy-clonic circulation off Newfoundland and diminished


trade winds in the subtropics. The midlatitude atmo-spheric alteration is speculated to be the surface sig-nature of forced Rossby waves initiated within the Trop-ics. This hypothesis is confirmed from model experi-ments using CCM3, where diabatic heating anomaliesdiagnosed from observations are imposed along theITCZ. Anomalous convection in the western tropicalbasin is responsible in CCM3 for the excitation of aforced wave train extending northeastward from theCaribbian basin toward Scandinavia. Modifications intrade wind intensity in the northern tropical Atlanticoccurs at the same time as a response to altered rainfallover the Sahel. The model teleconnection pattern re-produces fairly well the observed anomalous low-levelatmospheric flow related to HS. The associated turbulentand radiative fluxes at the sea surface would tend towarm up the eastern and subtropical basins and cool theregion east of Newfoundland, as observed. This allowsus to conclude that the HS can be viewed as the localimprint of tropically induced extratropical atmosphericchanges.

The late summer SST anomalies associated with theHS pattern were hypothesized to influence the atmo-spheric circulation over the North Atlantic and Europein late fall/early winter. To test this observational hy-pothesis, experiments with ARPEGE were performedwhere the observed evolution of the HS mode is pre-scribed from August to November in the model SSTboundary conditions. The model response is character-ized by a baroclinic structure through September withhigh (low) surface pressure above cold (warm) SSTanomalies, followed by an equivalent barotropic struc-ture from October onward that resembles the NAO. Thebaroclinic response can be interpreted as the atmospher-ic tendency to damp the extratropical SST anomaliesand may confirm the fact that the extratropical atmo-sphere by itself is not responsible for the HS. The bar-otropic response can be considered as the signature oflarge-scale atmospheric circulation in response to theextratropical SST anomalies. We have shown that anom-alous vertical shear collocated with the anomalous me-ridional SST gradient is associated with changes instorm activity starting in October. The alteration of thestorm track is hypothesized to be of central importanceto initiate and sustain large-scale barotropic anomaliesthrough eddy–mean flow interaction. The model seemsto be more receptive to the SST anomalies when thestorminess and associated upper-level jet are well de-veloped. This emphasizes the importance of the cli-matological background to explain the sign and the tim-ing of the atmospheric response to the persistent HSSST anomalies (see also Peng et al. 1995 but for latewinter).

b. Discussion

The analyses presented here from observations andfrom model experiments have insisted on the importance

of both summer tropical and extratropical conditions forpossibly influencing the next winter’s NAO. Tropicalatmospheric changes seem to initiate a bridge betweensummer and early winter, and the resulting extratropicalSST anomalies seems to be of importance as they persistenough to feed back to the atmosphere when the cli-matological background state is favorable. Therefore, itis important to consider together, in a global perspective,tropical and extratropical signals as they are not inde-pendent and contribute, by a cooperative and construc-tive interaction, to the variability of the winter NAO. Itwould be interesting, in particular, to test the ocean–atmosphere link presented in this paper from coupledmodel integrations.

In such a context, we can expand the present seasonalsummer-to-winter bridge and propose a winter-to-winterbridge to explain the year-to-year persistence of theNAO and part of its low frequency. Let us now startwith the winter–early spring season (from January tolate April) and a given phase of the NAO. As describedin the introduction, the prevailing forcing of the at-mosphere on the underlying ocean during the winter–early spring months imprints a tripolelike structure ofSST anomalies, which tends to be maximum in spring(Cayan 1992). Anomalous trade wind intensity peaks atthat time (Hastenrath and Greischar 1993) and tends tobuild or reinforce the so-called oceanic interhemisphericSST gradient in the deep Tropics. As detailed in Suttonet al. (2000), the atmospheric response to the oceaniccross-equatorial gradient is maximum for late spring/summer months (June–September) and is characterizedby a meridional displacement of the ITCZ toward thewarmer hemisphere. We have now established a linkbetween the summer tropical change and the previouswinter NAO conditions. Vimont et al. (2001) have di-agnosed a similar connection, but for the Pacific Ocean.Based on that link and on the present study on the re-lationship between summer SST conditions and the nextwinter’s NAO, the same NAO phase as the previouswinter would be favored and a new cycle would startwith the reappearance of the same-sign winter NorthAtlantic SST tripole. Summer months might thus appearvery important since, first, they are sensitive to the trop-ical SST anomalies forced to a great extent by the pre-vious winter–spring NAO and, second, they prepare theextratropical oceanic HS anomalies to which the at-mosphere is sensitive at the beginning of next winter.This temporal bridge over summer might explain partof the redenning of the winter NAO spectrum.

How can this bridge or cycle be interrupted? We haveshown that the key vector sustaining the global mech-anism is associated with the tropical convection overthe western part of the Atlantic basin and over the Sahelto a lesser extent. A simple analysis of variance (AN-OVA) shows that the internal atmospheric variabilitydominates in this region despite being located withinthe Tropics. It represents between 55% and 70% of thetotal variance depending on season, as estimated from


ARPEGE (Cassou and Terray 2001), and can thereforedisrupt the bridge. In addition to this internal compo-nent, El Nino–Southern Oscillation is one obvious can-didate to alter the Atlantic cycle as it remotely affectsthe strength and the position of the tropical convergencezones (Klein et al. 1999; Sutton et al. 2000; Gianniniet al. 2001; Alexander et al. 2002). The local interactionbetween the land and the atmosphere over the SouthAmerican continent might also play a role (Geldney andValdez 2000), as well as the Indian Ocean variabilityvia its effect on the branch of the Walker circulationover the African continent. The so-called ‘‘AtlanticNino’’ (Zebiak 1993; Czaja and Frankignoul 2002) isalso expected to modify the convection pattern of thetropical Atlantic and ultimately affects the midlatitudes(Drevillon et al. 2003).

Finally, the decadal variability of the subtropical andsubpolar gyres (Sutton and Allen 1997) might modulatethe efficiency of the remote tropical influences in im-printing extratropical SST anomalies, depending on thephase of the low frequency oceanic mode. Above all,we have to keep in mind that the internal dynamics ofthe extratropical North Atlantic atmosphere is by fardominant, even in summer, and can also create its ownoceanic surface anomalies that may affect the followingwinter NAO. A large amount of variance of the totalextratropical variability is controlled by internal atmo-spheric processes and the present study only explainsabout 15%–20% of the signal (based on the ARPEGEmodel response using ANOVA techniques, Cassou andTerray 2001).

Nevertheless, the lagged relationship between thesummer SST anomalies and the following winter at-mospheric circulation anomalies might be exploited forpredictability purposes (Rodwell and Folland 2002).The scenario presented here, however, may be too lim-ited due to the linear constraint on the analyses. Fol-lowing a nonlinear approach used in Cassou et al. (2004)based on cluster analysis, it appears that the negativephase of the NAO is indeed preferably linked to theprevious summer HS mode, whereas the positive phaseof the NAO is better associated with a different summerSST structure more reminiscent of the SST tripole. Itwould therefore be interesting to further investigate theasymmetry between the positive and negative phase ofthe NAO and their relationship with previous oceanicconditions.

Acknowledgments. This work is supported by the Eu-ropean Community via the 5th framework programPREDICATE project under Contract EVK2-CT-1999-00020 and by NOAA under Grant NA06GP0394. Theauthors want to thank A. Giannini, R. Sutton, J. Chiang,and K. Trenberth for stimulating discussions. The firstauthor is grateful to E. Maisonnave, S. Valcke, and C.Perigaud for their constant support and valuable cri-tiques and to A. Phillips for help accessing and plottingthe data with the NCL software.

REFERENCES

Alexander, M. A., I. Blade, M. Newman, J. R. Lanzante, N. C. Lau,and J. D. Scott, 2002: The atmospheric bridge: The influence ofENSO teleconnections on air–sea interaction over the globaloceans. J. Climate, 15, 2205–2231.

Barsugli, J., and D. S. Battisti, 1998: The basic effects of atmosphere–ocean thermal coupling on midlatitude variability. J. Atmos. Sci.,55, 477–493.

Battisti, D. S., U. S. Bhatt, and M. A. Alexander, 1995: A modelingstudy of the interannual variability in the wintertime North At-lantic Ocean. J. Climate, 8, 3067–3083.

Branstator, G., and S. E. Haupt, 1998: An empirical model of bar-otropic atmospheric dynamics and its response to tropical forc-ing. J. Climate, 11, 2645–2667.

Cassou, C., and L. Terray, 2001: Oceanic forcing of the wintertimelow-frequency atmospheric variability in the North Atlantic Eu-ropean sector: A study with the ARPEGE model. J. Climate, 14,4266–4291.

——, ——, J. W. Hurrell, and C. Deser, 2004: North Atlantic winterclimate regimes: Spatial asymmetry, stationarity with time, andoceanic forcing. J. Climate, 17, 1055–1068.

Cayan, D. R., 1992: Latent and sensible heat flux anomalies over thenorthern oceans: Driving the sea surface temperature. J. Phys.Oceanogr., 22, 859–881.

Czaja, A., and C. Frankignoul, 1999: Influence of the North AtlanticSST on the atmospheric circulation. Geophys. Res. Lett., 26,2969–2972.

——, and ——, 2002: Observed impact of Atlantic SST anomalieson the North Atlantic Oscillation. J. Climate, 15, 606–623.

de Coetlogon, G., and C. Frankignoul, 2003: Persistence of wintersea surface temperature in the North Atlantic. J. Climate, 16,1364–1377.

Deque, M., C. Dreveton, A. Braun, and D. Cariolle, 1994: The climateversion of Arpege/IFS: A contribution to the French communityclimate modelling. Climate Dyn., 10, 249–266.

Deser, C., and M. S. Timlin, 1997: Atmosphere–ocean interaction onweekly timescales in the North Atlantic and North Pacific. J.Climate, 10, 393–408.

——, M. A. Alexander, and M. S. Timlin, 2003: On the persistenceof sea surface temperature anomalies in midlatitudes. J. Climate,16, 57–72.

Doblas-Reyes, F. J., M. Deque, F. Valero, and D. B. Stephenson, 1998:North Atlantic wintertime intraseasonal variability and its sen-sitivity to GCM horizontal resolution. Tellus, 50A, 573–595.

——, M. A. Pastor, M. J. Casado, and M. Deque, 2001: Wintertimewestward-travelling planetary scale perturbations over the Euro–Atlantic region. Climate Dyn., 17, 811–824.

Drevillon, M., L. Terray, P. Rogel, and C. Cassou, 2001: MidlatitudeAtlantic SST influence on European winter climate variabilityin the NCEP–NCAR Reanalysis. Climate Dyn., 18, 331–344.

——, C. Cassou, and L. Terray, 2003: Model study of the wintertimeatmospheric response to fall tropical Atlantic SST anomalies.Quart. J. Roy. Meteor. Soc., 129, 2591–2611.

Fontaine, B., S. Janicot, and P. Roucou, 1999: Coupled ocean–at-mosphere surface variability and its climate impacts in the trop-ical Atlantic region. Climate Dyn., 15, 451–473.

Frankignoul, C., and K. Hasselmann, 1977: Stochastic climate mod-els. Part II: Application to sea surface temperature variabilityand thermocline variability. Tellus, 29, 284–305.

——, A. Czaja, and B. L’Heveder, 1998: Air–sea feedback in theNorth Atlantic and surface boundary conditions for ocean mod-els. J. Climate, 11, 2310–2324.

Gedney, N., and P. J. Valdez, 2000: The effect of Amazonian defor-estation of the Northern Hemisphere circulation and climate.Geophys. Res. Lett., 27, 3053–3056.

Giannini, A., M. A. Cane, and Y. Kushnir, 2001: Interdecadal changesin the ENSO teleconnection to the Caribbean region and theNorth Atlantic Oscillation. J. Climate, 14, 2867–2879.

——, R. Saravanan, and P. Chang, 2003: Oceanic forcing of Sahel


rainfall on interannual and interdecadal time scales. Science, 302,1027–1030.

Hastenrath, S., and L. Greischar, 1993: Circulation mechanisms re-lated to northeast Brazil rainfall anomalies. J. Geophys. Res.,98, 5093–5102.

Held, I. M., S. W. Lyons, and S. Nigam, 1989: Transients and theextratropical response to El Nino. J. Atmos. Sci., 46, 163–174.

Hoerling, M. P., J. W. Hurrell, and T. Xu, 2001: Tropical origins forNorth Atlantic climate change. Science, 292, 90–92.

Hoskins, B., 1987: Diagnostics of forced and free variability in theatmosphere. Atmospheric and Oceanic Variability, H. Cattle,Ed., Royal Meteorological Society, 57–74.

——, and R. Pierce, 1983: Large Scale Dynamical Processes in theAtmosphere. Academic Press, 397 pp.

——, and P. J. Valdez, 1990: On the existence of storm tracks. J.Atmos. Sci., 47, 1854–1864.

——, I. James, and G. White, 1983: The shape, propagation andmean flow interaction of large scale weather systems. J. Atmos.Sci., 40, 1595–1612.

Hulme, M., and M. G. New, 1997: Dependence of large-scale pre-cipitation climatologies on temporal and spatial sampling. J. Cli-mate, 10, 1099–1113.

Hurrell, J. W., 2003: An overview of the North Atlantic Oscillation.Geophys. Monogr. Series, No. 134, Amer. Geophys. Union, 1–22.

——, J. J. Hack, B. A. Boville, D. L. Williamson, and J. T. Kiehl,1998: The dynamical simulation of the NCAR Community Cli-mate Model version 3 (CCM3). J. Climate, 11, 1207–1236.

——, M. P. Hoerling, A. Phillips, and T. Xu, 2004: Twentieth centuryNorth Atlantic climate change. Part I: Assessing determinism.Climate Dyn., in press.

Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Re-analysis Project. Bull. Amer. Meteor. Soc., 77, 437–471.

Kiehl, J. T., J. J. Hack, G. B. Bonan, B. A. Boville, B. P. Briegleb,D. L. Williamson, and P. J. Rasch, 1996: Description of theNCAR Community Climate Model (CCM3). NCAR Tech. NoteNCAR TN-4201STR, 152 pp. [Available from NCAR, Boulder,CO 80307.]

Klein, S. A., B. J. Soden, and N. G. Lau, 1999: Remote sea surfacetemperature variations during ENSO: Evidence for a tropicalatmospheric bridge. J. Climate, 12, 917–932.

Kushnir, Y., W. A. Robinson, I. Blade, N. M. J. Hall, S. Peng, andR. T. Sutton, 2002: Atmospheric GCM response to extratropicalSST anomalies: Synthesis and evaluation. J. Climate, 15, 2233–2256.

Marshall, J., and Coauthors, 2001: Review: North Atlantic climatevariability: Phenomena, impacts and mechanisms. Int. J. Cli-matol., 21, 1863–1898.

Mehta, V. M., M. J. Suarez, J. V. Manganello, and T. Delworth, 2000:Oceanic influence on the North Atlantic Oscillation and asso-ciated Northern Hemisphere climate variations, 1959–1993.Geophys. Res. Lett., 27, 121–124.

Norris, J. R., Y. Zhuang, and J. W. Wallace, 1998: Role of low cloudsin summertime atmosphere–ocean interactions over the NorthPacific. J. Climate, 11, 2482–2490.

Okumura, Y., S. P. Xie, A. Numaguti, and Y. Tanimoto, 2001: TropicalAtlantic air–sea interaction and its influence on the NAO. Geo-phys. Res. Lett., 28, 1507–1510.

Peng, S., and J. S. Whitaker, 1999: Mechanisms determining theatmospheric response to midlatitude SST anomalies. J. Climate,12, 1393–1408.

——, and W. A. Robinson, 2001: Relationships between atmospheric

internal variability and the responses to an extratropical SSTanomaly. J. Climate, 14, 2943–2959.

——, L. A. Mysak, H. Ritchie, J. Derome, and B. Dugas, 1995: Thedifference between early and midwinter atmospheric responsesto sea surface temperature anomalies in the northwest Atlantic.J. Climate, 8, 137–157.

——, W. A. Robinson, and S. Li, 2003: Mechanisms for the NAOresponses to the North Atlantic SST tripole. J. Climate, 16,1987–2004.

Robinson, W. A., 2000: Review of WETS—The workshop on extra-tropical SST anomalies. Bull. Amer. Meteor. Soc., 81, 567–577.

Rodwell, M. J., and C. K. Folland, 2002: Atlantic air–sea interactionand seasonal predictability. Quart. J. Roy. Meteor. Soc., 128,1413–1443.

——, D. P. Rowell, and C. K. Folland, 1999: Oceanic forcing of thewintertime North Atlantic Oscillation and European climate. Na-ture, 398, 320–323.

Sardeshmukh, P. D., and B. J. Hoskins, 1988: The generation of globalrotational flow by steady idealized tropical divergence. J. Atmos.Sci., 45, 1228–1251.

Stephenson, D. B., H. Wanner, S. Br nnimann, and J. Luterbacher,ø2003: The history of scientific research on the North AtlanticOscillation. Geophys. Monogr., No. 134, Amer. Geophys. Union,37–50.

Sutton, R. T., and M. R. Allen, 1997: Decadal variability of NorthAtlantic sea surface temperature and climate. Nature, 388, 563–567.

——, and D. L. R. Hodson, 2003: The influence of the ocean onNorth Atlantic climate variability 1871–1999. J. Climate, 16,3269–3313.

——, S. P. Jewson, and D. P. Rowell, 2000: The elements of climatevariability in the tropical Atlantic region. J. Climate, 13, 3261–3284.

——, W. A. Norton, and S. P. Jewson, 2001: The North AtlanticOscillation—What role for the ocean? Atmos. Sci. Lett., 1, 89–100.

Terray, L., and C. Cassou, 2002: Tropical Atlantic sea surface tem-perature forcing of the quasi-decadal climate variability over theNorth Atlantic–Europe region. J. Climate, 15, 3170–3187.

Trenberth, K. E., 1986: An assessment of the impact of transienteddies on the zonal flow during a blocking episode using lo-calized Eliassen–Palm flux diagnostics. J. Atmos. Sci., 43, 2070–2087.

Vimont, D. J., D. S. Battisti, and A. C. Hirst, 2001: Footprinting: Aseasonal link between the midlatitudes and tropics. Geophys.Res. Lett., 28, 3923–3926.

Wallace, J. M., and D. Gutzler, 1981: Teleconnections in the geo-potentiel fields during the Northern Hemisphere winter. Mon.Wea. Rev., 109, 784–812.

Watanabe, M., and M. Kimoto, 1999: Tropical–extratropical connec-tion in the Atlantic atmosphere ocean variability. Geophys. Res.Lett., 26, 2247–2250.

——, and ——, 2000: Atmosphere–ocean thermal coupling in theNorth Atlantic: A positive feedback. Quart. J. Roy. Meteor. Soc.,126, 3343–3369.

Webster, P. J., 1982: Seasonality in the local and remote atmosphericresponse to sea surface temperature anomalies. J. Atmos. Sci.,39, 722–733.

Woodruff, S. D., H. F. Diaz, J. D. Elms, and S. J. Worley, 1998:COADS Release 2 data and metadata enhancements for im-provements of marine surface flux fields. Phys. Chem. Earth,23, 517–527.

Zebiak, S., 1993: Air–sea interaction in the equatorial Atlantic region.J. Climate, 6, 1567–1586.

summer sea surface temperature conditions in ... - cerfacs.fr · 1september 2004 cassou et al. 3349...

Documents